I need help with predicting an attribute

ramos213

Hallo all,

I have a dataset that contains the yearly gas use 2010- 2018, temperature from 2010 - 2018, the provinces in the country that i want to analyze, and the speed of wind. What i want to do know is to predict what the usage of gas will be in the future. When i observe the data set it is clearly that the usage of gas is getting less by year and that there is correlation between temperature and gas usage. I tried to get an decision tree but it wont work for some reason. 

Can someone help me with predicting the gas usage?

Thanks in advance

lionelderkrikor

Hi @ramos213,

Have you tried to connect the in port to the Generate Macro operator input port ?

If the issue is still present after doing this connection, please share your data and your process in order we can reproduce and understand your issue.

Regards,

Lionel

ramos213

It worked, thanks alot guys!

varunm1

Hello @ramos213

This seems like you are trying to forecast future usage of gas. If you don't have any future data without labels then you need to go for forecasting related to time series data. Here is a link that helps you understand how time series analysis works.

https://rapidminer.com/resource/time-series-analysis/

Do let us know if this helps. If not, please clearly inform, how your data looks and do you have any unlabelled data to predict and the way you are building models.

ramos213

Hey Varunm1,

I watched the video and it is what i need, but i keep getting an error when i try to forecast. I did everything exactly like the video shows but for some reason the macro doesnt give input to cross validation. 

varunm1

Hello @ramos213

The input of cross-validation is an "example set". I guess you are using a wrong input to cross-validation operator.

ramos213

Hey,

I forgot to connect it indeed, but know i have another problem.

varunm1

Hello @ramos213

Can you provide us with the data and your process? You can download your process by going to FILE --> Export Process and then attach here in the thread with data set.

ramos213

Oh, this what i have now 

ramos213

This is my process

ramos213

The dataset "NoordBrabant" is what i used in the process.

lionelderkrikor

@ramos213

You have only 9 examples in your initial example set and 4 examples after windowing !!!
Thus you can not perform a 10-folds cross -validation (there are no enough examples).
You can not build a reliable model with so few data : You have to increase significantly the size of your training set....

Regards,

Lionel

PS : if you want absolutly a working process, set the k-folds parameters of CV to k = 4.

ramos213

I first tried with this dataset, the one that i just posted is one of the 3 provinces that i wanted to analyze. So this dataset contains wat i just posted but for some reason i couldnt get it to work in windowing, i think it has to do with the colomn years.

lionelderkrikor

Yes , in deed the error is due to the fact that you have duplicate years values in your initial dataset.
You have 27 examples in your initial dataset for your 3 provinces, thus you have 9 examples for each  province after splitting according the provinces : 
That's you did and you did the right thing but ...I have to insist :  9 examples is no enough to build a relevant and reliable model.
Try to increase the size of your dataset by finding the variables values before 2010 (for example) .... 

Thanks you for your understanding,

Regards,

Lionel